Review for NeurIPS paper: Dirichlet Graph Variational Autoencoder

Although the validation section would benefit from a somewhat more extensive set of baselines, the paper presents a nice and apparently well-working extension to VAEs.